New Simulation Method Enables Early Testing of Social Robots Without Human Participants

A collaborative study by the University of Surrey and the University of Hamburg has introduced a groundbreaking method that allows researchers to test social robots without the need for human participants. Presented at the IEEE International Conference on Robotics and Automation (ICRA), the study aims to streamline early development by using simulations instead of real-time human trials. Researchers developed a dynamic scanpath prediction model that enables humanoid robots to simulate human eye movements, determining what a person would likely focus on in social settings.

The model was validated using two publicly available datasets and was shown to perform effectively even in noisy and unpredictable environments. This innovation has important implications for social robotics applications in fields like education, health care, and customer service, where human-like interaction is crucial.

According to Dr. Di Fu, co-lead of the study and lecturer in cognitive neuroscience, the approach allows researchers to verify if robots are paying attention to socially relevant cues without the immediate need for human testing. By matching real-world gaze data with simulated responses, the model can reliably evaluate social attention in realistic conditions. This shift in methodology could significantly reduce the cost and time required in the early stages of social robotics research, paving the way for more refined and scalable development in complex social scenarios.